Skip to content

A Discrete Fourier Transform (DFT), a Fast Wavelet Transform (FWT), and a Wavelet Packet Transform (WPT) algorithm in 1-D, 2-D, and 3-D using normalized orthogonal (orthonormal) Haar, Coiflet, Daubechie, Legendre and normalized biorthognal wavelets in Java.

master
Go to file
Code

Latest commit

 

Git stats

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
 
 
 
 
lib
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

README.md

JWave

Java library keeping orthogonal / orthonormal and bi-orthogonal wavelets

The continuous integration Build Status on travis-ci.com!

Introduction

Java implementation of a Discrete Fourier Transform (DFT), a Fast Wavelet Transform (FWT), and a Wavelet Packet Transform (WPT) algorithm. All algorithms are available in 1-D, 2-D, and 3-D. The wavelet transform algorithms are using normalized orthogonal or if available orthonormal wavelets. The comon wavelets like Haar, Coiflet, Daubechies, Symlets, and Legendre are available. Additionally there are also some Bi-Orthogonal and unusal wavelets implemented - in total around 50 wavelets.

The implementation of JWave is based on several software design patterns and - hopefully - appears therefore user-friendly.

GETTING STARTED

First steps

Have a look at the little how to down the page or try looking at the unit tests.

Can I perform odd samplings like, e.g. 127 data points?

However, the bare algorithms of JWave do only support data sampled by 2^p | p E N; e.g. 2, 4, 8, 16, .. 128, 256, 512, 1024, .. and so on. Please use the AncientEgyptianDecomposition class for odd samplings (most frequently asked question)! You can find it down the page.

Why are the results totally different then expected?!

Why do the results (hilbert spaces) look totally different to those from my matlab or some other implementation I found on net?!

In most cases, other libraries construct the orthogonal / orthonormal bases of the wavelet and scaling function in a different way. Especially for those bases of multiple dimension or wavelets of higher dimension, repectively. But does this hurt?

Totally not!

The why can be found in mathematics. Due to using some orthogonal transform (or better, an orthogonal base), it is up to oneself how to construct this base (as long it stays orthogonal over all dimensions). Next it is also up to oneself how to apply the sequence of the transform steps. Both does not influence any performance of the wavelet transforms! But again why?

The base stays orthogonal, and one's data is unit rotated and mirrowed differently, which makes a long story short.

Additionally the application of the transform - independent of using some different rotating and mirrowing base - is like dancing some 90's techno: As long as you do the same amount of steps independently of the performed sequence, even in 2-D and 3-D dimensions, the expected magic wavelets can bring in, will be there, and stays the same! For example, the result in values, and e.g the compression rates will stay exactely the same. Only all intermediate performed results (or intermediate hilbert spaces) will be different, if someone else dances differently to otherones.

HowTo

For a quick test, pull the repository and then: ant && ant test. This builds a JWave.jar and the corresponding unit tests. Afterwards all units test are executed.

Doing own stuff e.g. data compression

For a lot of own stuff with JWave, have a look at the main junit test file / method: a lot of examples!

For example, how to perform a (losless) data compression with over 98 % compression rate using all available wavelets is shown by the following junit test!

Some easy code ..

example for 1-D DFT:

Transform t = new Transform( new DiscreteFourierTransform( ) );

// arrTime = { r1, c1, r2, c2, ... }`
double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };

double[ ] arrFreq = t.forward( arrTime ); // 1-D DFT forward

double[ ] arrReco = t.reverse( arrFreq ); // 1-D DFT reverse

example for 1-D, 2-D FWT:

Transform t = new Transform( new FastWaveletTransform( new Haar1( ) ) );

double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };

double[ ] arrHilb = t.forward( arrTime ); // 1-D FWT Haar forward

double[ ] arrReco = t.reverse( arrHilb ); // 1-D FWT Haar reverse

double[ ][ ] matTime = { { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. } };

double[ ][ ] matHilb = t.forward( matTime ); // 2-D FWT Haar forward

double[ ][ ] matReco = t.reverse( matHilb ); // 2-D FWT Haar reverse
// example in 3-D in common to 2-D using a N^3 double[ ][ ][ ] space.

example for 1-D, 2-D WPT:

Transform t = new Transform( new WaveletPacketTransform( new Haar1( ) ) );

double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1., 1. };

double[ ] arrHilb = t.forward( arrTime ); // 1-D WPT Haar forward

double[ ] arrReco = t.reverse( arrHilb ); // 1-D WPT Haar reverse

double[ ][ ] matTime = { { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. },
                         { 1., 1., 1., 1. } };

double[ ][ ] matHilb = t.forward( matTime ); // 2-D WPT Haar forward

double[ ][ ] matReco = t.reverse( matHilb ); // 2-D WPT Haar reverse

// example in 3-D in common to 2-D using a N^3 double[ ][ ][ ] space.

example for 1-D FWT of arbitrary length:

Transform t = new Transform(
               new AncientEgyptianDecomposition(
                new FastWaveletTransform(
                 new Haar1( ) ) ) );

double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1. }; // length = 7

double[ ] arrHilb = t.forward( arrTime ); // 1-D AED FWT Haar forward

//           |    2 steps    |   1 step   | 0  |
// arrHilb = { 2., 0., 0., 0., 1.41421, 0., 1. };
double[ ] arrReco = t.reverse( arrHilb ); // 1-D AED FWT Haar reverse

example for 1-D WPT (WPD) of arbitrary length:

Transform t = new Transform(
               new AncientEgyptianDecomposition(
                new WaveletPacketTransform(
                 new Haar1( ) ) ) );

double[ ] arrTime = { 1., 1., 1., 1., 1., 1., 1. }; // length = 7

double[ ] arrHilb = t.forward( arrTime ); // 1-D AED WPT Haar forward

double[ ] arrReco = t.reverse( arrHilb ); // 1-D AED WPT Haar reverse

Have fun! :-)

CONTACT

If there are doubts, try mailing me, otherwise have fun with JWave.

LICENSE

JWave is distributed under the MIT License (MIT); this file is part of.

Copyright (c) 2008-2020 Christian (graetz23@gmail.com)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

VERSION

JWave is in version 200303.

CHANGE LOG

version 200303:

  • updating copyright and contact information to graetz23

version 180222:

  • updating project description, especially using markdown for the README.md,
  • updated JUnit imports and travis-ci.org configuration for continuous integration,

version 160218:

  • added a new type of wavelet transform algorithm: Shifting Wavelet Transform.
  • the algorithm shifts a wavelet by smallest length of 2 over the input array,
  • then by the double of 4, 8, 16, .., p-1, p.
  • the reverse transform takes the largest wavelength of p and shifts,
  • then by half of p-1, .., 16, 8, 4, 2.

version 160109:

  • moved the JUnit tests to an own source directory
  • updated the build.xml file
  • automatically run all files with regex **/*Test* in build directory
  • updated the build.xml - allowing for OS specific builds now
  • set JUnit path OS dependent; MAC, WIN, or GNU/Linux and Unix
  • additionally set for WIN OS org.hamcrest.core JAR

version 160107:

  • added junit test for compressing a sine signal:
  • sine signal of 1024 * 1024 samples by 1024 oscillations
  • calculating the compression rate in percent; e.g. 99.70703125 % by Daubechies 20 wavelet
  • calculating the absolute maximal differences; e.g. 3.086649707542688E-4 by Daubechies 20 wavelet
  • added method for calculating compression rate to class Compressor

version 160106:

  • added build.xml for using ant
  • set in build.xml your path JUnit4, e.g. /usr/share/java/junit.jar
  • ! to have the build done:
  • ant
  • should work easy peasy
  • ! for a quick console example:
  • ant run
  • ! for running the junit tests
  • ant test
  • ! to have a run or test on your system:
  • java -cp ./dist/JWave.jar jwave.JWave Fast Wavelet Transform Daubechies 20
  • java -cp /usr/share/java/junit4.jar:./dist/JWave.jar org.junit.runner.JUnitCore jwave.TransformTest
  • added automatic build using travis-ci.com
  • ! https://travis-ci.org/graetz23/JWave
  • have a look at ./.travis.yml
  • added README.md
  • added LICENSE.md
  • updated Copyright to the years 2008-2018
  • fixed bug in JWave.java for calling console example with Haar wavelet
You can’t perform that action at this time.